DIY Robotics: Building Your Own AI Robot

Object tracking in Computer Vision is the task of automatically identifying objects in an image.

There are good tutorials on how to use Microsoft or Google’s apis to do object tracking on a picture or video frame. Most if not all use a picture taken by a photographer a few years in the past. A dated picture is not very useful for object tracking. 

The following picture is from a Tutorial teaching how to do object tracking with Microsoft Cognitive Services.

In this example a kitchen appliance has been located at 730,66. Those coordinates are relative to the photographer who took the picture possibly years ago . Is that kitchen appliance there now? Probably not. Do you care?

However, If we use object tracking on a picture taken a few milliseconds previously on a robot we can use that data more effectively. The following picture frame was taken from a robot.

In real-time it records and applies object tracking frame by frame. In this example it identified a cat (on a picture card) while it was driving.

Perhaps you are not interested in cats. Maybe you want to train your robot to detect your neighbor’s dog. Or maybe you want to find Bigfoot. Object tracking is a wonderful technology. But where and how it is applied is very important. I am just saying old pictures are not interesting. Make it real-time.

Leave a Reply

Your email address will not be published. Required fields are marked *